The dual-resonant enhancement of mechanical and optical response in cavity optomechanical magnetometers enables precision sensing of magnetic fields. In previous working prototypes of such magnetometers, a cavity optomechanical system is functionalized by manually epoxy-bonding a grain of magnetostrictive material. While this approach allows proof-of-principle demonstrations, practical applications require more scalable and reproducible fabrication pathways. In this work, we developed a multiple-step method to scalably fabricate optomechanical magnetometers on a silicon chip, with reproducible performance across different devices. The key step is to develop a process to sputter coat a magnetostrictive film onto high quality toroidal microresonators, without degradation of the optical quality factor. A peak sensitivity of 585 pT/Hz is achieved, which is comparable with previously reported results using epoxy-bonding. Furthermore, we demonstrate that thermally annealing the sputtered film can improve the magnetometer sensitivity by a factor of 6.3.

Ultrasensitive magnetometers are key sensing components for various applications, such as magnetic anomaly detection,1,2 mineral exploration,3,4 magnetoencephalography,5,6 and magnetic resonance imaging.7,8 Currently, the most advanced magnetometer is the superconducting quantum interference device (SQUID) based magnetometer.9–12 However, the requirement of cryogenic cooling increases the complexity of SQUID magnetometer systems. To circumvent this requirement, various high precision magnetometers operating at room temperature have been developed in the last decade, such as atomic magnetometers,13,14 nitrogen vacancy center magnetometers,15–20 and cavity optomechanical magnetometers.21–28 Among them, cavity optomechanical magnetometers offer the advantages of small size, weight, and power consumption; ease of on-chip integration; high sensitivity; and broad bandwidth. For instance, highly sensitive magnetometry has been realized by incorporating a magnetostrictive material with a whispering gallery mode (WGM) microcavity,21–25 while torque magnetometry has been demonstrated by integrating a high-magnetization material with a WGM microcavity26,27 or with a nanoscale photonic crystal cavity.28 In magnetostriction-based cavity optomechanical magnetometers, the strain induced by a magnetic field on the embedded magnetostrictive material deforms the optical cavity. The resulting cavity resonance shift is read out optically. The combination of resonance-enhanced mechanical and optical response29–31 enables unprecedented transduction sensitivity,32,33 surpassing that of the previously demonstrated magnetostrictive magnetometers34–37 by several orders of magnitude.

The first working prototype of a cavity optomechanical magnetometer was realized by manually depositing a grain of magnetostrictive material on top of a microcavity and affixing it using epoxy.21 This magnetometer achieved a sensitivity of hundreds of nT/Hz. The sensitivity was further improved to a level of hundreds of pT/Hz, by fabricating a central hole in the cavity structure and depositing the magnetostrictive material into it.22 However, the manual deposition process requires precise positioning of micro-sized grains relative to the microcavity. Combined with the use of epoxy-bonding, this makes the approach ill-suited for scalable fabrication. Furthermore, both optimization of the overlap of the magnetostriction to mechanical motion and reproducible performance across devices are hard to realize due to the random geometry, orientation, and size of the magnetostrictive material grain in each device. To overcome these challenges, in this work, we develop a controllable fabrication method, which involves deterministically sputter coating thin films onto the microcavities. While developed here for magnetometry, the ability to sputter coat materials onto high quality (Q) optical cavities has broader applications, providing a versatile functionalization method, both for sensors and actuators. This is challenging for two reasons. On the one hand, high Q cavities are very sensitive to losses induced by materials deposited in the vicinity of the optical field. On the other hand, the laser reflow process,38 which melts the periphery of the cavity to smooth the surface, can affect the properties of the deposited materials. Here we resolve these issues, inspired by previous work where thermal evaporation39 or electron beam evaporation40 was used to deposit metal structures. Our approach uses a multiple-step process to protect the optical cavity from the sputter coated material and the material from laser reflow. This allows parallel sputter coating of functional magnetostrictive films onto arrays of microcavities, without observed loss in optical quality. We characterize the sensitivities of ten devices, showing quite similar performances across devices. A peak sensitivity of 585 pT/Hz is achieved, which is comparable with previously reported devices fabricated using the manual deposition method.22 Furthermore, by thermally annealing the sputtered magnetostrictive film, we show that the magnetostrictive coefficient is increased by a factor of 6.3, leading to improvement in the sensitivity. This sputter coating method provides a scalable and reproducible fabrication pathway for microcavity optomechanical magnetometers on a silicon chip.

Figure 1(a) shows a schematic side view of the designed magnetometer. It consists of a silica microtoroid38,41 with a disk of magnetostrictive material embedded inside, supported by a silicon pedestal. The silica toroid supports high Q WGMs that confine the optical field near its perimeter. The system also supports high Q mechanical modes, e.g., radial breathing mode [Fig. 1(b)]. The mechanical motion modifies the circumference of the optical cavity and thus shifts the optical resonance. An external magnetic field induces a stress tensor T(r) in the magnetostrictive material which generates a body force with density of f(r)=T(r).42 This induces an effective driving force F on the mechanical modes of the device which depends on the overlap between f(r) and the mechanical mode profile u(r), F=f(r)u(r)dV,21 modulating the resonance frequencies of the WGMs of the device, which can be read out optically with high precision.

FIG. 1.

(a) Schematic of the side view of the designed magnetometer, which consists of a silica microtoroid with a magnetostrictive material embedded inside, supported by a silicon pedestal. R1 and R2 denote the outer radius of the toroid and radius of the magnetostrictive disk, respectively. (b) The radial breathing mode profile of the disk [in the dotted region in (a)] along the radial direction. The gray border shows the equilibrium position of the disk. (c) The displacement distribution of the radial breathing mode [shown in (b)], normalized to its maximum value. (d) R2u(R2) as a function of the magnetostrictive disk radius R2, normalized to its maximum value (where R2 = R1 = 30 µm). The red dotted line is for R2 = 23 µm, with the value reaching 80% of its maximum value.

FIG. 1.

(a) Schematic of the side view of the designed magnetometer, which consists of a silica microtoroid with a magnetostrictive material embedded inside, supported by a silicon pedestal. R1 and R2 denote the outer radius of the toroid and radius of the magnetostrictive disk, respectively. (b) The radial breathing mode profile of the disk [in the dotted region in (a)] along the radial direction. The gray border shows the equilibrium position of the disk. (c) The displacement distribution of the radial breathing mode [shown in (b)], normalized to its maximum value. (d) R2u(R2) as a function of the magnetostrictive disk radius R2, normalized to its maximum value (where R2 = R1 = 30 µm). The red dotted line is for R2 = 23 µm, with the value reaching 80% of its maximum value.

Close modal

In contrast to previous work,21,22 fine control of the fabrication process allows the geometry of the magnetometer to be designed to optimize the overlap between f(r) and u(r) and therefore the magnetic field sensitivity. Here we consider a cavity geometry with a toroid outer radius of R1 = 30 µm and vary the radius of the magnetostrictive disk R2 to optimize the overlap. The disk thickness is chosen to be t = 2 µm, and the radius of the silicon pedestal is chosen to be 5 µm to minimize the constraint that it imposes on the expansion of the magnetostrictive film, while remaining large enough to support the disk. We simulate the mechanical mode profiles of the resonator through COMSOL Multiphysics. We optimize specifically for the radial breathing mode since its motion is primarily radial and therefore well overlapped with the direction of the magnetostrictive expansion, and it has the strongest coupling to the frequencies of the WGM resonances of the device and therefore the largest optical transduction sensitivity. As this mode is axis-symmetric along the y axis [perpendicular to the plane of the disk, as denoted in Fig. 1(a)], we can simplify u(r) into u(r)r^, where u(r) is the amplitude of the mechanical displacement and r^ is the unit vector along the radial direction in the plane of the disk. The cross-sectional profile of the radial breathing mode is shown in Fig. 1(b), and its displacement distribution u(r) normalized to its maximum value is plotted in Fig. 1(c).

A homogeneous magnetic field oriented in the x direction [the horizontal direction, as denoted in Fig. 1(a)] applied to the magnetostrictive material causes the material to stretch in the same direction. The magnetostriction induced stress tensor has only one single component Txx = , generating a body force along the x direction f(r)=f(r)x^=Bα(r)xx^, in which B is the magnetic field strength, α is the magnetostrictive coefficient, and x^ is the unit vector along the x direction. As α(r)x is only nonzero at the interface (r = R2) between the magnetostrictive material and the silica,21,42 the body force amplitude becomes f(r)=Bαδ(rR2)cosβ, with β being the angle between the body force direction x^ and the mechanical displacement direction r^. Transforming the integral into cylindrical coordinates (r, β, h), and inserting r^x^=cosβ and dV = rdβdhdr, the effective driving force becomes F=0tdh0R1Bαδ(rR2)u(r)rdr02πcos2βdβ=πBαtR2u(R2).21 To explore the dependence of F on the magnetostrictive disk radius, we plot R2u(R2) (normalized to its maximum value) as a function of R2 in Fig. 1(d). It can be seen that R2u(R2) increases monotonically with R2 and reaches its maximum when R2 = R1 = 30 µm, suggesting better magnetic field sensitivity for a larger magnetostrictive disk radius. Taking into account the experimental fabrication constraints, with a toroid minor diameter of 5 µm required to achieve desired optical Q and a 2 µm separation between the toroid and the magnetostrictive material to protect the magnetostrictive material during reflow, the largest feasible magnetostrictive disk radius is R2 = 23 µm. For this choice of the magnetostrictive disk radius, the overlap reaches 80% of its maximum value.

The magnetometer fabrication process is shown in Fig. 2. Starting from a silicon wafer (thickness of 500 µm) with a 2 μm-thick thermally oxidized silica layer on top [shown in (a)], silica annular disks are patterned with standard photolithography using a positive photoresist AZ 1518, followed by a hydrofluoric (HF) acid etching process [step (b)]. Next a thin film of magnetostrictive material with similar thickness as the silica is sputter coated into the hole in each annular disk. We choose Terfenol-D (Fe2Tb0.3Dy0.7) as the magnetostrictive material, as this material has high magnetostriction at room temperature and can be deposited via sputter coating.43 Before sputter coating, the negative photoresist AZ 2070 is spin coated over the wafer, with the photoresist over the holes removed using a second photolithography step [step (c)]. This defines the region for Terfenol-D deposition while protecting the perimeter of each disk from the sputter coated material. In this step, the radius of the opening in the photoresist is designed to be 5 µm larger than the holes on each side, to leave some tolerance for alignment in the photolithography process. This also leaves some overlap between Terfenol-D and silica to increase the bonding between them. Then a layer of 2 μm-thick Terfenol-D film is sputter coated on top of the wafer [step (d)]. Note that a 10 nm-thick gold layer is coated both below and above the Terfenol-D to protect it from oxidization; oxidization can spoil the magnetostriction and degrade the sensitivity of the magnetometer.43 The excess Terfenol-D and gold outside the holes are removed through a lift-off process using acetone [step (e)]. A xenon difluoride (XeF2) etch is then performed to undercut the silicon pedestal [step (f)], followed by a carbon dioxide (CO2) laser reflow process to melt the edge of the silica disk to form a toroid structure and increase the optical Q factor of the silica microcavity [step (g)]. At this stage, the depth of the undercut is chosen to both allow reflow of the rim of the silica disks into high quality microtoroids and to ensure a good thermal contact between the Terfenol-D film and the silicon pedestal. To reflow the silica disks, the temperature of the rim of the silica disks must be raised above its melting point of ∼1500 °C. However, as silicon has much larger thermal conductivity than silica, the silicon pedestal acts as an effective heat sink to transfer heat from the silica disk into the substrate38 and therefore to protect the Terfenol-D film from CO2 laser heating which can oxidize, ablate, or change the crystalline structure of the film. Finally a second XeF2 etch is performed to further underetch the silicon pedestal [step (h)], in order to release the Terfenol-D from the silicon pedestal for better magnetic actuation.

FIG. 2.

Schematic of the fabrication process for on-chip scalable cavity optomechanical magnetometers. [(a) and (b)] A silica annulus is fabricated through photolithography and HF etching. [(c)–(e)] A disk consisting of stacked layers of Au/Terfenol-D/Au thin films is embedded into the hole of the silica annulus through a second-step photolithography, electron-beam evaporation (for Au coating) and sputter coating (for Terfenol-D), and lift-off. [(f)–(h)] A XeF2 etch, CO2 laser reflow, and a second-step XeF2 etch are performed to undercut the silicon pedestal, make the silica toroid, and further undercut the silicon pedestal.

FIG. 2.

Schematic of the fabrication process for on-chip scalable cavity optomechanical magnetometers. [(a) and (b)] A silica annulus is fabricated through photolithography and HF etching. [(c)–(e)] A disk consisting of stacked layers of Au/Terfenol-D/Au thin films is embedded into the hole of the silica annulus through a second-step photolithography, electron-beam evaporation (for Au coating) and sputter coating (for Terfenol-D), and lift-off. [(f)–(h)] A XeF2 etch, CO2 laser reflow, and a second-step XeF2 etch are performed to undercut the silicon pedestal, make the silica toroid, and further undercut the silicon pedestal.

Close modal

A scanning electron microscope (SEM) image of a silicon chip with an array of silica microdisks with Terfenol-D disks embedded inside is shown in Fig. 3(a). A magnified image of one disk is shown in (b), exhibiting good uniformity of the Terfenol-D film. The thickness of the Terfenol-D film is measured to be ∼2.2 µm. The final magnetometer device is shown in Fig. 3(c), with a toroid diameter of ∼80 µm.

FIG. 3.

Scanning electron microscope (SEM) images of the fabricated devices. (a) A silicon chip with an array of silica disks embedded with Terfenol-D. (b) Zoomed-in view of one of the silica disks (prior to the first XeF2 etching process) with Terfenol-D embedded in the center. (c) The final device: a toroid microresonator with a Terfenol-D disk embedded in the center.

FIG. 3.

Scanning electron microscope (SEM) images of the fabricated devices. (a) A silicon chip with an array of silica disks embedded with Terfenol-D. (b) Zoomed-in view of one of the silica disks (prior to the first XeF2 etching process) with Terfenol-D embedded in the center. (c) The final device: a toroid microresonator with a Terfenol-D disk embedded in the center.

Close modal

To investigate the optical properties of the fabricated devices, we couple light from a laser in the 1550 nm wavelength band into the microtoroid through a tapered fiber with a diameter of ∼1 µm.44 The transmitted light is detected using a 125 MHz-bandwidth InGaAs photoreceiver. The wavelength of the light from the laser is swept to find the optical resonances. Typical optical Q factors are found to be in the range of 107–108, consistent with quality factors obtained in pure silica toroids.38,41 Figure 4 shows a transmission spectrum of the toroid from 1550 nm to 1560 nm wavelength range, in which multiple optical resonances appear, with a free spectral range (FSR) of ∼6.1 nm. A zoomed-in view of the highest-Q mode of this particular device is shown in the inset, with a linewidth of ∼0.03 pm, corresponding to an intrinsic optical Q of ∼5 × 107. The modes are identified through COMSOL Multiphysics simulation, with the cross-sectional field distribution of modes 1-6 shown in the top panel of Fig. 4.

FIG. 4.

Transmission spectrum of a toroid from 1550 nm to 1560 nm, with a free spectral range (FSR) of ∼6.1 nm. The cross-sectional optical field distribution of modes 1-6 are shown in the top panel, obtained through COMSOL Multiphysics simulations. The azimuthal numbers are identified to be m = 243, 233, 239, 232, 229, and 235 for modes 1–6 in the wavelength range between 1555 and 1560 nm. A zoomed-in view of the highest-Q mode of this particular device is shown in the inset, with the black curve being the measured data and the red dashed curve being the Lorentzian fitting of the resonance dip. The mode has a linewidth of ∼0.03 pm, corresponding to an intrinsic optical Q of ∼5 × 107.

FIG. 4.

Transmission spectrum of a toroid from 1550 nm to 1560 nm, with a free spectral range (FSR) of ∼6.1 nm. The cross-sectional optical field distribution of modes 1-6 are shown in the top panel, obtained through COMSOL Multiphysics simulations. The azimuthal numbers are identified to be m = 243, 233, 239, 232, 229, and 235 for modes 1–6 in the wavelength range between 1555 and 1560 nm. A zoomed-in view of the highest-Q mode of this particular device is shown in the inset, with the black curve being the measured data and the red dashed curve being the Lorentzian fitting of the resonance dip. The mode has a linewidth of ∼0.03 pm, corresponding to an intrinsic optical Q of ∼5 × 107.

Close modal

The mechanical motion of the microresonator translates into a periodic modulation of the intracavity field. Intensity (phase) modulation can be optically read out by locking the frequency of the probe light on the side (center) of the cavity resonance. In our experiment, the detuning of the laser is stabilized by thermally locking the laser at half maximum of the transmission on the blue-detuned side of an optical resonance.45 The thermal locking works as follows: when the laser light is coupled into the cavity, it heats the cavity and induces a cavity resonance shift due to the thermal-optic effect. When the laser is locked on the blue detuned side of the cavity mode, if the laser wavelength drifts to a longer (shorter) wavelength, the intracavity power increases (decreases) as the detuning decreases (increases). This causes the cavity resonance to red-shift (blue-shift) due to the positive thermal-optic coefficient of silica,46 maintaining the detuning between the laser wavelength and the cavity mode unchanged.

The output probe light, captured by the tapered fiber, is detected by using a photoreceiver. The direct current (DC) part of the photocurrent from the photoreceiver is sent to an oscilloscope to monitor the transmission. The alternating current (AC) part is sent to an electronic spectrum analyzer (ESA) to measure the mechanical spectrum of the cavity. To test the magnetic response of the magnetometer, a coil is used to produce a magnetic field with a known strength. The frequency of the produced magnetic field can be swept by sweeping the frequency of the voltage applied to the coil from the output port of an electronic network analyzer (ENA), and the magnetic response of the magnetometer at each frequency is measured with the input port of the same ENA.

In a cavity optomechanical magnetometer, the main noise sources consist of the thermal noise Nth from the environment and the shot noise Ns from the probe laser. The signal to noise ratio SNR induced by a magnetic field is given by SNR = S/(Nth + Ns). As both S and Nth are proportional to the probe power, and Ns is proportional to the square-root of the probe power, increasing the probe power reduces the relative contribution from shot noise and therefore provides a better SNR and magnetic field sensitivity.24 At sufficiently high probe power, the sensitivity will saturate at the thermal-noise-limited sensitivity. In our experiment, the thermal noise dominant regime is reached for most of the measured frequency ranges (0.5-60 MHz), with a probe power of only ∼50 µW. Figure 5 shows the noise power spectrum of one of the fabricated magnetometers in the frequency range from 1.5 MHz to 12.5 MHz. Multiple peaks can be seen in the spectrum, which correspond to thermally excited mechanical modes. These modes are identified through a finite element method simulation using COMSOL Multiphysics, with mode profiles of five of the mechanical modes shown in the top panel. A zoomed-in view of the noise power spectrum in the frequency range of 8 MHz-8.8 MHz is shown in the black curve in Fig. 6(a), fitting well to a pair of Lorentzian peaks (red curve). The two modes correspond to two originally degenerated crown modes, with the same mode profile but spatially rotated by 30° relative to each other, such that the nodes of one mode correspond to the antinodes of the other. We expect that the small frequency splitting between these two modes is caused by the slight asymmetry in the structure induced in the fabrication process.

FIG. 5.

Noise power spectrum of a magnetometer, with the peaks representing the thermally excited mechanical modes. The profiles of five of the modes are shown in the top panel, obtained using COMSOL Multiphysics. The measured/simulated frequencies for the five modes are 2.11 MHz/2.14 MHz, 4.91 MHz/4.97 MHz, 8.24 MHz/8.30 MHz, 8.45 MHz/8.32 MHz, and 11.90 MHz/12.40 MHz.

FIG. 5.

Noise power spectrum of a magnetometer, with the peaks representing the thermally excited mechanical modes. The profiles of five of the modes are shown in the top panel, obtained using COMSOL Multiphysics. The measured/simulated frequencies for the five modes are 2.11 MHz/2.14 MHz, 4.91 MHz/4.97 MHz, 8.24 MHz/8.30 MHz, 8.45 MHz/8.32 MHz, and 11.90 MHz/12.40 MHz.

Close modal
FIG. 6.

(a) Zoomed-in view of the noise power spectrum (black curve) in the frequency range of 8 MHz-8.8 MHz (the blue shaded region in Fig. 5), with the red curve representing the Lorentzian fitting of the two modes. Inset: Noise power spectrum of the magnetometer driven with a known magnetic field at a reference frequency of 8.48 MHz, from which the signal to noise ratio (SNR) and thus the sensitivity at this frequency can be derived. (b) Magnetic response (black curve) of the magnetometer in the frequency range of 8.0 MHz-8.8 MHz. The blue dashed curve is the envelope of the magnetic response spectrum, following the mechanical resonance spectrum. (c) Sensitivity spectrum of the magnetometer in the frequency range of 8 MHz-8.8 MHz. (d) Zoomed-in view of the sensitivity spectrum in the frequency range of 8.465 MHz-8.502 MHz, with the peak sensitivity of ∼585 pT/Hz and the 3 dB bandwidth of 16.8 kHz.

FIG. 6.

(a) Zoomed-in view of the noise power spectrum (black curve) in the frequency range of 8 MHz-8.8 MHz (the blue shaded region in Fig. 5), with the red curve representing the Lorentzian fitting of the two modes. Inset: Noise power spectrum of the magnetometer driven with a known magnetic field at a reference frequency of 8.48 MHz, from which the signal to noise ratio (SNR) and thus the sensitivity at this frequency can be derived. (b) Magnetic response (black curve) of the magnetometer in the frequency range of 8.0 MHz-8.8 MHz. The blue dashed curve is the envelope of the magnetic response spectrum, following the mechanical resonance spectrum. (c) Sensitivity spectrum of the magnetometer in the frequency range of 8 MHz-8.8 MHz. (d) Zoomed-in view of the sensitivity spectrum in the frequency range of 8.465 MHz-8.502 MHz, with the peak sensitivity of ∼585 pT/Hz and the 3 dB bandwidth of 16.8 kHz.

Close modal

The magnetic field sensitivity of the magnetometer is then characterized. A magnetic field with known strength Bref at a certain frequency ωref is first applied to the magnetometer. The inset of Fig. 6(a) shows the noise power spectrum of the magnetometer driven with a magnetic field at 8.48 MHz, from which the SNR, and thus the sensitivity, δB(ωref) can be obtained. The sensitivity is given by δB(ωref)=Bref/SNR×RBW, with RBW being the resolution bandwidth of the ESA. The SNR is found to scale linearly with the strength of Bref in the range up to 100 µT, which is consistent with previous results.22,23 In order to obtain the magnetic response across the full frequency range, the frequency of the magnetic field is swept using the ENA. The magnetic response R(ω) in the frequency range from DC to 90 MHz is measured. The thermal-noise-limited sensitivity25 is reached in the frequency range from 0.5 MHz to 60 MHz. The sensitivity degrades in the frequency range below 0.5 MHz due to the increased laser noise and above 60 MHz due to the decreased mechanical response. The magnetic response above 90 MHz is not measured due to the limitation of the strength of the test magnetic field generated from the coil. With the noise power spectrum N(ω), the sensitivity δB(ω) of the magnetometer in this frequency range can be derived from δB(ω)=δB(ωref)[N(ω)R(ωref)]/[N(ωref)R(ω)].21 The magnetic field response is found to vary significantly over frequency ranges of ∼10 kHz, with these features remaining stationary with time, as shown in the magnetic response in the frequency range from 8.0 MHz to 8.8 MHz [black curve in Fig. 6(b)]. This is consistent with previously reported results and is attributed to sharp magnetostrictive resonances.21 Modulo these steep resonances, the envelope of the magnetic response as a function of frequency follows the mechanical resonance spectrum, as shown by the blue dashed curve in Fig. 6(b). The derived sensitivity in this frequency range is shown in Fig. 6(c). A zoomed-in view of the sensitivity spectrum in the frequency range of 8.465 MHz-8.502 MHz is shown in Fig. 6(d), with a peak sensitivity of ∼585 pT/Hz and a 3 dB bandwidth of 16.8 kHz.

Due to the parallel and reproducible fabrication method, it is found that all magnetometers fabricated in this way show quite similar sensitivities. In Fig. 7(a), we plot the peak sensitivity of ten different magnetometers fabricated on the same wafer, with sensitivity values ranging from 0.585 nT/Hz to 7 nT/Hz. The mean sensitivity achieved is 3.65 nT/Hz (the dotted line) with a standard deviation of 2.19 nT/Hz (the shaded area). As a comparison, in our experience, the sensitivities of magnetometers fabricated with manual Terfenol-D deposition method in previous studies can vary by more than two orders of magnitude.

FIG. 7.

(a) Peak sensitivities of 10 different magnetometers, showing reproducible sensitivities. The dotted line and the shaded area represent the mean and standard deviation. (b) The noise power spectra for a magnetometer before (red curve) and after (blue curve) thermal annealing, driven with a magnetic field at a frequency of 20.85 MHz.

FIG. 7.

(a) Peak sensitivities of 10 different magnetometers, showing reproducible sensitivities. The dotted line and the shaded area represent the mean and standard deviation. (b) The noise power spectra for a magnetometer before (red curve) and after (blue curve) thermal annealing, driven with a magnetic field at a frequency of 20.85 MHz.

Close modal

It has been shown that sputter coated Terfenol-D is amorphous which results in an order of magnitude smaller magnetostrictive coefficient than crystalline Terfenol-D.47 Two methods could be used to crystallize the Terfenol-D film: increasing the temperature of the substrate to 400 °C during the sputter coating process or post-annealing the amorphous Terfenol-D film at 400 °C.47 In our case, the first method is not applicable, as the photoresist on the wafer would not survive 400 °C. We therefore perform post-annealing at 400 °C for 6 h at high vacuum (∼10−7 Torr) to avoid oxidization. For devices with the optimal silicon pedestal size with a diameter of ∼10 µm, as described earlier, we find that the large mismatch in the thermal coefficients of Terfenol-D (11 × 10−6 K−1)43 and silica (0.55 × 10−6 K−1)46 causes the microtoroids to crack during annealing. Instead, we test thermal annealing on an alternative magnetometer with a larger pedestal with a diameter of ∼50 µm. Although devices with such large pedestals are expected to have relatively poor sensitivities due to the constraint that the pedestal imposes on the expansion of Terfenol-D, they serve as good candidates to test the effect of thermal annealing on the magnetostriction of the sputter coated Terfenol-D film. In this case, the silicon pedestal (with a thermal expansion coefficient of 4 × 10−6 K−1)48 reduces the thermal expansion of the Terfenol-D film during the annealing process and allows the process to be performed without cracking. Besides crystallizing the sputtered Terfenol-D film, we note that thermal annealing also releases the stress in the film, which may also affect the magnetostriction.47 Figure 7(b) shows the noise power spectra of the magnetometer before (red curve) and after (blue curve) thermal annealing, driven by a magnetic field with a known strength at a frequency of 20.85 MHz. The SNR is improved by 16 dB by annealing, corresponding to a sensitivity enhancement by a factor of 101.6/2 = 6.3, from 7.5 nT/Hz to 1.2 nT/Hz.

Studies show that the magnetostrictive coefficient of Terfenol-D is very sensitive to its composition.43 In our experiment, the atomic composition of the sputter coated Terfenol-D film is measured via energy dispersive X-ray spectroscopy (EDS), to be 74% Fe, 8% Tb, and 18% Dy, which deviates from the ideal composition: 66% Fe, 10% Tb, and 24% Dy. Optimization of the composition through modification of the sputter coating parameters is likely to improve the sensitivity, possibly to a level comparable with the state-of-the-art microscale magnetometers.9,12,16 The sensitivity can be further improved by increasing the sensing area of the magnetometer,21,25 potentially to reach that of commercially available magnetometers such as flux-gate49 and SQUID11 magnetometers.

In summary, we report the development of high sensitivity optomechanical magnetometers that can be scalably fabricated on a silicon chip. To achieve this, we have developed a new process to functionalize high-Q optical microcavities with a magnetostrictive material, through sputter coating, without causing degradation of the optical Q or the magnetostrictive performance. This controllable and deterministic fabrication method also allows the device geometry to be designed to optimize magnetic field sensitivities and ensures reproducible sensitivities across devices. A peak sensitivity of 585 pT/Hz is achieved, comparable to the previously reported result using manual Terfenol-D deposition. It is also demonstrated that thermally annealing the sputter coated Terfenol-D film can improve the magnetostrictive coefficient and therefore the sensitivity of the magnetometer, by a factor of 6.3. This work opens up possibilities for a range of applications that benefit from high sensitivity without cryogenics, such as on-chip microfluidic nuclear magnetic resonance, magnetoencephalography, and geological survey.

We thank James Bennett, Hamish Greenall, Brian Vyhnalek, and Felix Miranda for the helpful discussions. We acknowledge support from the DARPA QuARSAR Program, Australian Research Council projects (Nos. DP140100734 and FT140100650), and Australian Defence Science and Technology Group projects (Nos. CERA49 and CERA50). Bei-Bei Li also acknowledges the support from the University of Queensland Postdoctoral Research Fellowship (No. 2014001447). With the exception of sputter coating and laser reflow, device fabrication was performed within the Queensland Node of the Australian Nanofabrication Facility.

1.
C.
Li
,
S.
Huang
,
D.
Wei
,
Y.
Zhong
, and
K. Y.
Gong
,
J. Eng. Sci. Tech. Rev.
8
,
105
(
2015
).
2.
A.
Sheinker
,
L.
Frumkis
,
B.
Ginzburg
,
N.
Salomonski
, and
B.-Z.
Kaplan
,
IEEE Trans. Magn.
45
,
160
(
2009
).
3.
A.
Edelstein
,
J. Phys.: Condens. Matter
19
,
165217
(
2007
).
4.
H. G.
Meyer
,
R.
Stolz
,
A.
Chwala
, and
M.
Schulz
,
Phys. Status Solidi
2
,
1504
(
2005
).
5.
E.
Boto
,
N.
Holmes
,
J.
Leggett
,
G.
Roberts
,
V.
Shah
,
S. S.
Meyer
,
L. D.
Muoz
,
K. J.
Mullinger
,
T. M.
Tierney
,
S.
Bestmann
,
G. R.
Barnes
,
R.
Bowtell
, and
M. J.
Brookes
,
Nature
555
,
657
(
2018
).
6.
H.
Xia
,
A. B.-A.
Baranga
,
D.
Hoffman
, and
M. V.
Romalis
,
Appl. Phys. Lett.
89
,
211104
(
2006
).
7.
D. R.
Glenn
,
D. B.
Bucher
,
J.
Lee
,
M. D.
Lukin
,
H.
Park
, and
R. L.
Walsworth
,
Nature
555
,
351
(
2018
).
8.
I.
Savukov
and
T.
Karaulanov
,
Appl. Phys. Lett.
103
,
043703
(
2013
).
9.
J. R.
Kirtley
,
M. B.
Ketchen
,
K. G.
Stawiasz
,
J. Z.
Sun
,
W. J.
Gallagher
,
S. H.
Blanton
, and
S. J.
Wind
,
Appl. Phys. Lett.
66
,
1138
(
1995
).
10.
F.
Baudenbacher
,
L. E.
Fong
,
J. R.
Holzer
, and
M.
Radparvar
,
Appl. Phys. Lett.
82
,
3487
(
2003
).
11.
M. I.
Faley
,
U.
Poppe
,
K.
Urban
,
D. N.
Paulson
, and
R. L.
Fagaly
,
J. Phys.: Conf. Ser.
43
,
1199
(
2006
).
12.
J. R.
Kirtley
,
L.
Paulius
,
A. J.
Rosenberg
,
J. C.
Palmstrom
,
C. M.
Holland
,
E. M.
Spanton
,
D.
Schiessl
,
C. L.
Jermain
,
J.
Gibbons
,
Y.-K.-K.
Fung
,
M. E.
Huber
,
D. C.
Ralph
,
M. B.
Ketchen
,
G. W.
Gibson
, Jr.
, and
K. A.
Moler
,
Rev. Sci. Instrum.
87
,
093702
(
2016
).
13.
H. B.
Dang
,
A. C.
Maloof
, and
M. V.
Romalis
,
Appl. Phys. Lett.
97
,
151110
(
2010
).
14.
M.
Vengalattore
,
J. M.
Higbie
,
S. R.
Leslie
,
J.
Guzman
,
L. E.
Sadler
, and
D. M.
Stamper-Kurn
,
Phys. Rev. Lett.
98
,
200801
(
2007
).
15.
G.
Balasubramanian
,
P.
Neumann
,
D.
Twitchen
,
M.
Markham
,
R.
Kolesov
,
N.
Mizuochi
,
J.
Isoya
,
J.
Achard
,
J.
Beck
,
J.
Tissler
,
V.
Jacques
,
P. R.
Hemmer
,
F.
Jelezko
, and
J.
Wrachtrup
,
Nat. Mater.
8
,
383
(
2009
).
16.
T.
Wolf
,
P.
Neumann
,
K.
Nakamura
,
H.
Sumiya
,
T.
Ohshima
,
J.
Isoya
, and
J.
Wrachtrup
,
Phys. Rev. X
5
,
041001
(
2015
).
17.
V. M.
Acosta
,
E.
Bauch
,
A.
Jarmola
,
L. J.
Zipp
,
M. P.
Ledbetter
, and
D.
Budker
,
Appl. Phys. Lett.
97
,
174104
(
2010
).
18.
L. M.
Pham
,
D. L.
Sage
,
P. L.
Stanwix
,
T. K.
Yeung
,
D.
Glenn
,
A.
Trifonov
,
P.
Cappellaro
,
P. R.
Hemmer
,
M. D.
Lukin
,
H.
Park
,
A.
Yacoby
, and
R. L.
Walsworth
,
New J. Phys.
13
,
045021
(
2011
).
19.
D.
Le Sage
,
L. M.
Pham
,
N.
Bar-Gill
,
C.
Belthangady
,
M. D.
Lukin
,
A.
Yacoby
, and
R. L.
Walsworth
,
Phys. Rev. B
85
,
121202(R)
(
2012
).
20.
K.
Jensen
,
N.
Leefer
,
A.
Jarmola
,
Y.
Dumeige
,
V. M.
Acosta
,
P.
Kehayias
,
B.
Patton
, and
D.
Budker
,
Phys. Rev. Lett.
112
,
160802
(
2014
).
21.
S.
Forstner
,
S.
Prams
,
J.
Knittel
,
E. D.
vanOoijen
,
J. D.
Swaim
,
G. I.
Harris
,
A.
Szorkovszky
,
W. P.
Bowen
, and
H.
Rubinsztein-Dunlop
,
Phys. Rev. Lett.
108
,
120801
(
2012
).
22.
S.
Forstner
,
E.
Sheridan
,
J.
Knittel
,
C. L.
Humphreys
,
G. A.
Brawley
,
H.
Rubinsztein-Dunlop
, and
W. P.
Bowen
,
Adv. Mater.
26
,
6348
(
2014
).
23.
C.
Yu
,
J.
Janousek
,
E.
Sheridan
,
D. L.
McAuslan
,
H.
Rubinsztein-Dunlop
,
P. K.
Lam
,
Y.
Zhang
, and
W. P.
Bowen
,
Phys. Rev. Appl.
5
,
044007
(
2016
).
24.
B.-B.
Li
,
J.
Bilek
,
U. B.
Hoff
,
L. S.
Madsen
,
S.
Forstner
,
V.
Prakash
,
C.
Schafermeier
,
T.
Gehring
,
W. P.
Bowen
, and
U. L.
Andersen
,
Optica
5
,
850
(
2018
).
25.
W.
Bowen
and
C.
Yu
, “
Cavity optomechanical magnetometry
,” in
High Sensitivity Magnetometers
, Volume 19 of the Series “Smart Sensors, Measurement and Instrumentation” (
Springer
,
2016
).
26.
J.
Zhu
,
G.
Zhao
,
I.
Savukov
, and
L.
Yang
,
Sci. Rep.
7
,
8896
(
2017
).
27.
E.
Freeman
,
C.-Y.
Wang
,
V.
Sumaria
,
S. J.
Schiff
,
Z.
Liu
, and
S.
Tadigadapa
,
AIP Adv.
8
,
065214
(
2018
).
28.
M.
Wu
,
N. L.-Y.
Wu
,
T.
Firdous
,
F. F.
Sani
,
J. E.
Losby
,
M. R.
Freeman
, and
P. E.
Barclay
,
Nat. Nanotechnol.
12
,
127
(
2017
).
29.
T. J.
Kippenberg
and
K. J.
Vahala
,
Science
321
,
1172
(
2008
).
30.
M.
Aspelmeyer
,
T. J.
Kippenberg
, and
F.
Marquardt
,
Rev. Mod. Phys.
86
,
1391
(
2014
).
31.
W. P.
Bowen
and
G.
Milburn
,
Quantum Optomechanics
(
CRC Press
,
Boca Raton
,
2016
).
32.
E.
Gavartin
,
P.
Verlot
, and
T. J.
Kippenberg
,
Nat. Nanotechnol.
7
,
509
(
2012
).
33.
K. H.
Lee
,
T. G.
McRae
,
G. I.
Harris
,
J.
Knittel
, and
W. P.
Bowen
,
Phys. Rev. Lett.
104
,
123604
(
2010
).
34.
F.
Bucholtz
,
D. M.
Dagenais
, and
K. P.
Koo
,
Electron. Lett.
25
,
1719
(
1989
).
35.
R.
Osiander
,
S. A.
Ecelberger
,
R. B.
Givens
,
D. K.
Wickenden
,
J. C.
Murphy
, and
T. J.
Kistenmacher
,
Appl. Phys. Lett.
69
,
2930
(
1996
).
36.
S.
Dong
,
J.-F.
Li
, and
D.
Viehland
,
Appl. Phys. Lett.
83
,
2265
(
2003
).
37.
Y.
Hui
,
T.
Nan
,
N. X.
Sun
, and
M.
Rinaldi
,
J. Microelectromech. Syst.
24
,
134
(
2014
).
38.
D. K.
Armani
,
T. J.
Kippenberg
,
S. M.
Spillane
, and
K. J.
Vahala
,
Nature
42
,
925
(
2003
).
39.
D.
Armani
,
B.
Min
,
A.
Martin
, and
K. J.
Vahala
,
Appl. Phys. Lett.
85
,
5439
(
2004
).
40.
C. G.
Baker
,
C.
Bekker
,
D. L.
Mcauslan
,
E.
Sheridan
, and
W. P.
Bowen
,
Opt. Express
24
,
20400
(
2016
).
41.
X.
Jiang
,
L.
Shao
,
S.-X.
Zhang
,
X.
Yi
,
J.
Wiersig
,
L.
Wang
,
Q.
Gong
,
M.
Loncar
,
L.
Yang
, and
Y.-F.
Xiao
,
Science
358
,
344
(
2017
).
42.
A. N.
Cleland
,
Foundations of Nanomechanics: From Solid-State Theory to Device Applications
(
Springer
,
New York
,
2003
).
43.
G.
Engdahl
,
Handbook of Giant Magnetostrictive Materials
(
Academic Press
,
2000
).
44.
J. C.
Knight
,
G.
Cheung
,
F.
Jacques
, and
T. A.
Birks
,
Opt. Lett.
22
,
1129
(
1997
).
45.
T. G.
McRae
,
K. H.
Lee
,
M.
McGovern
,
D.
Gwyther
, and
W. P.
Bowen
,
Opt. Express
17
,
21977
(
2009
).
46.
B.-B.
Li
,
Q.-Y.
Wang
,
Y.-F.
Xiao
,
X.-F.
Jiang
,
Y.
Li
,
L.
Xiao
, and
Q.
Gong
,
Appl. Phys. Lett.
96
,
251109
(
2010
).
47.
K. P.
Mohanchandra
,
S. V.
Prikhodko
,
K. P.
Wetzlar
,
W. Y.
Sun
,
P.
Nordeen
, and
G. P.
Carman
,
AIP Adv.
5
,
097119
(
2015
).
48.
H.
Watanabe
,
N.
Yamada
, and
M.
Okaji
,
Int. J. Thermophys.
25
,
221
(
2004
).
49.
R. H.
Koch
,
J. G.
Deak
, and
G.
Grinstein
,
Appl. Phys. Lett.
75
,
3862
(
1999
).